1 00:00:01,080 --> 00:00:04,000 Hello, welcome back to Motlop into session. 2 00:00:04,020 --> 00:00:09,210 We are going to learn how to save data to work in space for a new beginning. 3 00:00:09,240 --> 00:00:21,030 That's a story generating some data X equals two in a space again from zero to two PI, just like what 4 00:00:21,030 --> 00:00:21,960 we did before. 5 00:00:22,360 --> 00:00:30,660 And then I want 50 samples because that's what is giving us the best result for a designer's function 6 00:00:30,960 --> 00:00:35,550 and volumes equal to sign as a function of X. 7 00:00:36,180 --> 00:00:39,060 Finally, plot X and Y. 8 00:00:42,700 --> 00:00:46,270 OK, and called for an F to. 9 00:00:48,890 --> 00:00:49,580 Here we are. 10 00:00:49,610 --> 00:00:59,200 Let's just quickly choose your inputs X and Y, they are in the Matrix column and OK, that's fine. 11 00:00:59,240 --> 00:01:00,830 Let's just move on next. 12 00:01:01,340 --> 00:01:05,810 Three layers would be nice if we also calculated the formula for a dollar. 13 00:01:05,830 --> 00:01:11,830 You can just check it out and find a best number of hidden nephrons here in this example. 14 00:01:11,840 --> 00:01:15,170 It's just three and then click on next. 15 00:01:17,460 --> 00:01:18,570 Click on train. 16 00:01:22,130 --> 00:01:23,430 OK, here. 17 00:01:23,930 --> 00:01:27,320 Oh, that was interesting, actually, let me just show it to you again. 18 00:01:27,710 --> 00:01:35,820 This time the gradient is stopped over training and if you want to see the feet, let's check it very 19 00:01:35,840 --> 00:01:36,530 quickly. 20 00:01:36,740 --> 00:01:38,120 OK, that's a very good fit. 21 00:01:38,870 --> 00:01:40,890 Now, we don't need this one. 22 00:01:40,910 --> 00:01:43,420 Let me close it and click on next. 23 00:01:43,850 --> 00:01:45,410 And this is fine. 24 00:01:45,410 --> 00:01:47,200 Let's click on Next. 25 00:01:47,210 --> 00:01:50,660 And here we are, application development. 26 00:01:50,660 --> 00:01:52,100 Here we have different parts. 27 00:01:52,100 --> 00:01:53,780 We have Matlab function. 28 00:01:54,170 --> 00:01:59,060 If we want to generate some code, you can use Matlab, Imitrex only function. 29 00:01:59,060 --> 00:02:02,750 And here we have Simelane deployment. 30 00:02:03,890 --> 00:02:07,940 There are also some explanation about each which you can read. 31 00:02:08,150 --> 00:02:09,860 But let's just click on next. 32 00:02:09,860 --> 00:02:13,770 And here we are saved data to work space. 33 00:02:14,240 --> 00:02:19,490 We can save network to matlab network object and we can give you the name the default. 34 00:02:19,490 --> 00:02:20,330 Here is Nant. 35 00:02:20,330 --> 00:02:24,530 You can just choose any any name based on your data. 36 00:02:24,530 --> 00:02:29,120 You can save your performance and dataset any information. 37 00:02:29,420 --> 00:02:32,480 You can save the output of Matrix. 38 00:02:33,320 --> 00:02:41,930 You can save the error and inputs and target and more or all of them and you can just give the name 39 00:02:41,930 --> 00:02:42,550 result. 40 00:02:42,860 --> 00:02:46,190 But let me just I'll take them here. 41 00:02:46,850 --> 00:02:53,510 Let's see what would happen if we save the network to Matlab, the network object name Nant. 42 00:02:54,770 --> 00:03:02,090 Look at this workspace in this park you can only see X and Y my input in my output. 43 00:03:02,630 --> 00:03:10,940 But if I click on save result here, I shall see some output, some other output which is net. 44 00:03:11,210 --> 00:03:13,010 So here is our object net. 45 00:03:13,970 --> 00:03:19,600 If you double click on this net, you can see some information, Nancy. 46 00:03:19,640 --> 00:03:25,580 The object with different features here you can see the way the bias values and more. 47 00:03:25,850 --> 00:03:29,720 Oh, let me close this one in back to overcome in window. 48 00:03:30,230 --> 00:03:32,060 But what can we do with this net? 49 00:03:32,780 --> 00:03:37,280 Do use net to approximate a scientist, for example. 50 00:03:38,540 --> 00:03:46,850 I'm going to right here, sinus of Pi over to we all know what's the answer to answer this one. 51 00:03:47,450 --> 00:03:55,040 Now let's try to approximate the output using this net function net and then in parentheses we have 52 00:03:55,310 --> 00:04:00,890 pi over to OK, it's one point zero zero zero one. 53 00:04:00,890 --> 00:04:01,940 It's almost one. 54 00:04:02,360 --> 00:04:12,620 It shows that we train our network very well and approximate output of net function is almost the same 55 00:04:12,620 --> 00:04:16,370 with our Sina's function, which is one. 56 00:04:17,270 --> 00:04:18,860 We can do different things here. 57 00:04:18,860 --> 00:04:25,700 We can define y hat as nets function net of X. 58 00:04:28,040 --> 00:04:32,000 Here is the data of my hand, let me just call it again. 59 00:04:32,000 --> 00:04:35,510 But this time puts that semicolon. 60 00:04:35,810 --> 00:04:43,400 So boy hat is the output for my net function which we trained already with training between neurons 61 00:04:43,400 --> 00:04:46,220 in a hidden layer and fifty samples. 62 00:04:48,470 --> 00:04:55,700 Let's just compare the output of our Sina's function Y and the output of our neural network, which 63 00:04:55,700 --> 00:04:59,010 is here net, and we just put it on a Y hat. 64 00:04:59,570 --> 00:05:07,070 So I'm going to plot X and Y and I want to show it with the color of black. 65 00:05:07,460 --> 00:05:11,180 You can put K for the color of black. 66 00:05:11,180 --> 00:05:21,410 And then here I'm going to show X and each time Y hat the output of our neural network and I'm going 67 00:05:21,410 --> 00:05:24,200 to show it reads color for it. 68 00:05:26,660 --> 00:05:32,380 But let's just make one of them dotted line so we can compare them better. 69 00:05:32,660 --> 00:05:36,770 If you put it here than it would D-line, this red line would be. 70 00:05:39,860 --> 00:05:41,990 OK, zoom and see. 71 00:05:44,250 --> 00:05:48,730 We shall see two two lines. 72 00:05:49,400 --> 00:05:56,600 They are very close and that shows that our train network has a very good output. 73 00:05:57,350 --> 00:06:05,890 This dotted line is the output of over y hat of our neural network and this black line is forever Sina's. 74 00:06:06,650 --> 00:06:08,280 OK, it was a very good feat. 75 00:06:08,480 --> 00:06:10,130 Now close it. 76 00:06:10,310 --> 00:06:20,060 We can even show the errors, define another variable name and e I want to e to showed errors between 77 00:06:20,060 --> 00:06:21,200 why and why hat. 78 00:06:25,430 --> 00:06:27,140 You can also plot. 79 00:06:29,140 --> 00:06:31,270 X and E. 80 00:06:36,070 --> 00:06:46,600 OK, securities Netty is a function with different objects and features back to over fitting tool and 81 00:06:47,140 --> 00:06:49,240 check out this simple a script.